Monday, July 27, 2015

The PCSK9 enzyme, LDL cholesterol, and cardiovascular diseases


Cardiovascular diseases are currently the leading cause of death in most developed countries. They are particularly common among seniors; i.e., those aged 65 and older. Part of the reason for this is that infectious diseases do not kill as many people as they used to.

Given the trend toward population aging, with seniors making up an increasingly larger percentage of the population, the market for drugs against cardiovascular diseases is growing. A new class of such drugs is making the news lately; they target the PCSK9 enzyme ().

Enzymes are (usually) proteins that speed up chemical reactions, and are needed in virtually all metabolic processes that occur in cells. Proprotein convertase subtilisin/kexin type 9 (PCSK9) is an enzyme that degrades LDL cholesterol receptors on the surface of liver cells. Fewer LDL cholesterol receptors mean reduced uptake of the particles that carry LDL cholesterol, and thus more LDL particles in circulation. This may be problematic if these are small-dense LDL particles ().

Small-dense LDL particles include particles that are significantly smaller than the gaps in the endothelium (). The endothelium is a thin layer of cells that line the interior of arteries. Those gaps are about 25-26 nanometers (nm) in diameter. Small-dense LDL particles can contribute a lot more to the formation of atheromas (atherosclerotic plaques) in predisposed individuals than large-buoyant LDL particles.

There is evidence of the natural occurrence of low LDL cholesterol in individuals of African descent due to genetic mutations influencing PCSK9 levels (). This leads us to a very important question. By reducing PCSK9 in circulation, can we also reduce the incidence of cardiovascular disease?

The answer to this question depends on whether LDL cholesterol is a causative factor in cardiovascular disease. If it is, then reducing PCSK9 in circulation can indeed reduce the incidence of cardiovascular disease. The problem is that, most of the evidence so far suggests that LDL cholesterol is NOT a causative factor in cardiovascular disease.

Yes, there are studies that show that LDL cholesterol is correlated with cardiovascular disease, but the problem is that LDL cholesterol is a marker of other factors that are better candidates for causes of cardiovascular disease – hence the correlation. For example, LDL cholesterol goes up with mental stress (), and chronic mental stress seems to be a good candidate for a cause of cardiovascular disease.

LDL cholesterol is also a marker of a diet with more saturated fat in it (). In many contexts, a diet with more saturated fat in it is a more nutritious diet, which leads to a negative association between LDL cholesterol and mortality.

The graph below shows the shape of the association between total cholesterol (TOTCHOL) and mortality from all cardiovascular diseases (MVASC), based on an analysis of the China Study II dataset (). LDL cholesterol is the main component of total cholesterol in most people. The values are provided in standardized format; e.g., 0 is the average, 1 is one standard deviation above the mean, and so on. The best-fitting curve was obtained with the software WarpPLS ().

In fact, when we combine the totality of the evidence linking LDL cholesterol and cardiovascular diseases, LDL cholesterol seems to come out as a marker of protective factors. A reflection of this is a widely cited study by Weverling-Rijnsburger and colleagues, of LDL and HDL cholesterol as factors in cardiovascular diseases among people aged 85 and older (). The conclusions of the study were that:

- There was no association between LDL cholesterol level and risk of fatal cardiovascular disease.

- A low HDL cholesterol level was associated with a two-fold higher risk of fatal cardiovascular disease.

- Both low LDL cholesterol and low HDL cholesterol levels were associated with an increased mortality risk from infections.

The results above are particularly interesting because the study participants, given their ages, were at a high risk of mortality from cardiovascular diseases. It seems that the best scenario for these folks would have been a concomitant increase in both LDL and HDL cholesterol levels, which seems to be exactly what happens when one increases his or her intake of foods rich in saturated fat and dietary cholesterol ()!

Should you take a drug that targets the PCSK9 enzyme, to reduce your LDL cholesterol? Maybe you should ask Peter ().

21 comments:

Anonymous said...

Thanks for your thoughts.

But then, If not LDL what about LDL-P as several like Dayspring and Attia submit as the "driving" force of CVD?

Ned Kock said...

Hi Anon. That’s one option. Another is the IMT test:

http://healthcorrelator.blogspot.com/2011/05/interview-with-jimmy-moore-and-basics.html

http://healthcorrelator.blogspot.com/2012/09/familial-hypercholesteromia-why-rely-on.html

Ned Kock said...

By the way, here is a discussion on a LDL-P test (perhaps the most widely used):

http://healthcorrelator.blogspot.com/2012/10/the-anatomy-of-vap-test-report.html

cellulite exercises said...

Such a great and informative post you have shared here on health really useful and interesting. Thanks and keep sharing more blogs like this.

dearieme said...

I am unpersuaded by that curve as a meaningful best fit. It's clearly a lousy fit for the data on the right, say for abscissa > 1.5, because only one point falls beneath the line and about eleven above. The central part may indeed be well fitted by a curve or straight line that descends slightly to the right, but the fact that the curve shoots up as one moves one's eye towards the left extreme may be an artefact of just one outlying data point.

Ned Kock said...

Hi dearieme. That is the best-fitting “S curve”. Assuming that the relationship may be spurious – what do you think is the underlying reason for its shape?

dearieme said...

"That is the best-fitting “S curve”." Then the S-curve isn't well suited to the data. A plot of residuals would emphasise that.

Ned Kock said...

What do you mean by: “A plot of residuals would emphasize that”?

dearieme said...

The residuals - the differences between the fitted values and the actual data, plotted versus the abscissa variable, would show distinctly non-random behaviour at the right edge. They would probably look more random in the middle range: they would also look strongly non-random on the left extreme. This pattern supports the proposition that the curve-type that is being forced through the data is unsuitable.

How many parameters are there in an S-curve? if you fitted the data with a straight line, and then did an F-test to compare the fit of the S-curve and the line, would the S-curve prove superior?

Ned Kock said...

You are referring to a test of endogeneity; i.e., a test to check whether there is a significant correlation between a criterion’s residual and its predictor. A plot is not a good way of conducting such a test. Endogeneity is better tested by calculating the correlations among residuals and variables, which WarpPLS does automatically. This is not an issue here in this model.

Another possibility is controlling for the effect of a possible confounder, which is relatively easy to do in WarpPLS – if we know what to control for:

http://warppls.blogspot.com/2011/08/using-warppls-in-e-collaboration.html

http://warppls.blogspot.com/2010/02/how-do-i-control-for-effects-of-one-or.html

This is why I asked earlier: Assuming that the relationship may be spurious – what do you think is the underlying reason for its shape?

Confounders can lead to outliers, which may indeed distort relationships, by, for example, making linear relationships look like they are somewhat nonlinear.

One example of confounder leading to outliers is provided in the post linked below. The confounder is “proportion of seniors in the population (population age)”, and the analysis controls for its effect.

http://healthcorrelator.blogspot.com/2013/11/latitude-and-cancer-rates-in-us-states.html

dearieme said...

"what do you think is the underlying reason for its shape?": I have no idea. But I can spot a lousy correlation when one is presented.

Ned Kock said...

The nonlinear coefficient of association is significant, and so is the corresponding linear coefficient. The nonlinear function has a better fit than the linear. As you know from the original post (which is linked), the coefficients of association (which could be seen as adjusted correlations) were calculated through a multivariate analysis that already controlled for: SexM1F2 = sex, with 1 assigned to males and 2 to females; HDLCHOL = HDL cholesterol; and MSCHIST = mortality from schistosomiasis infection.

Btw, most academic health-related research studies test the effects of drug-based interventions on important health variables, such as mortality from this or that condition. Such research usually arrives at coefficients of association that are minute compared to the ones obtained for this model. And yet, those coefficients are reported as indications of significant effects; statistical significance is indeed achieved, but primarily due to the large sample sizes used and NOT due to magnitude of the effects. The vast majority of academic health-related research does NOT report effect sizes, which tend to be very small.

Maliha Sazin said...
This comment has been removed by a blog administrator.
Anonymous said...
This comment has been removed by a blog administrator.
hhaniya said...
This comment has been removed by a blog administrator.
how to remove cellulite said...
This comment has been removed by a blog administrator.
James Ringo said...

Hi Ned,

Important topic and medical research community has not done well with the cholesterol story. still, a medscape report http://www.medscape.com/viewarticle/841504 says "that high-risk patients with hypercholesterolemia who received the PCKS9 monoclonal antibody alirocumab (Sanofi/Regeneron Pharmaceuticals) plus statins for up to 78 weeks had consistently lowered LDL-cholesterol levels compared with those receiving placebo plus statins. Post hoc analysis also showed significantly lower rates of adverse cardiac events."

I don't follow the field at all closely, but if the cardiac effect, not just the lowered LDL-Chol effect holds up then would you need to re-think? This of course, a random assignment, placebo controlled trial so ancillary correlation seems unlikely.

Unknown said...
This comment has been removed by a blog administrator.
Mr Gbengz said...
This comment has been removed by a blog administrator.
Ned Kock said...

Spam comments above deleted.

Ned Kock said...

Hi James.

For non-FH individuals, much of the evidence suggests that LDL cholesterol and CVD are unrelated.

For individuals who suffer from FH, regular monitoring via intima-media thickness tests (which are not invasive) can lead to more effective treatments.

Atheromas are not made primarily of cholesterol; minerals are key components:

http://www.ncbi.nlm.nih.gov/pubmed/24134634